Hello, I have bought a new computer with I7-9700 and RTX 2080 SUPER for machine learning purposes. But after Tensorflow and its requirements installation (TensorFlow'u pip ile kurun), the old laptop with GTX 860M has been faster than new PC. (I executed same program at both machines) Could you point me to the right direction on what can be wrong? Should I use CUDA 10.0 instead of CUDA 10.1 and if yes, is RTX 2080 SUPER compatible with CUDA 10.0??
Thank you in advance
If you are installing tensorflow via pip package, you will indeed need cuda 10.0.
If you are running on Linux, you may also want to checkout the NVIDIA NGC Docker containers for TensorFlow. https://www.nvidia.com/en-us/gpu-cloud/containers/ These come with the latest toolkits and cudnn libraries which are continuously being improved for the latest GPUs.